fileformer
search⌘KtranslateENexpand_moreupload_fileConvert
Converters8 file families · 44 direct conversions
What changes in a conversionarrow_forward
tools / image-to-text

Image to text

Image to text OCR that copies the words out of a photo, screenshot or scan. Recognition runs in this tab with the open-source Tesseract engine. Every word is boxed on the image and coloured by how sure the engine was, so you know which ones to check.

Test receiptEnglishDeutschFrançaisEspañolItalianoloading engine…
confidentunsurecheck thisClick a word to select it in the text.
words…
avg confidence…
to check…
characters…
content_copyCopy textdownload.txt

Words to check

Below 70% confidence, lowest first. Numbers and names are worth a second look even when confident.
Reading…

Clean up the image, then compare

Tesseract reads dark text on a light, even background best. Turn on any of these, run again, and each run is added to the table so you can see whether it actually helped this image.

check_box_outline_blankGreyscaleRemoves colour so coloured paper or tinted photos don’t confuse the binariser.
check_box_outline_blankThreshold (Otsu)Turns every pixel black or white at the level that best splits ink from paper. Helps faded or uneven scans.
check_box_outline_blankEnlarge 2×Small text gets more pixels per letter. Helps screenshots and distant photos; slower to read.
refreshRun again with these
runwordsconfidenceto checktime
first run in progress…
Text size in this image…
010 px20 px50 px
Measured from the word boxes once reading finishes.

Line by line

Reading…

Export

Plain textLines + confidenceWords CSV
What’s in the editor above, including your corrections.
Reading…
downloadDownload receipt.txt

The engine behind it

1985Started at HP Labs in Bristol as a research OCR engine.
1995Among the top engines in the UNLV OCR accuracy test.
2005HP releases Tesseract as open source.
2006Development moves to Google sponsorship.
2018Tesseract 4 adds a neural-network (LSTM) line recogniser.
2021Tesseract 5.0 released; tesseract.js brings it to browsers via WebAssembly.